In [20]:
#!/usr/bin/env python
"""learn lime"""
Out[20]:
'learn lime'
In [21]:
# package
# from __future__ import print_function
import lime
import sklearn
import numpy as np
import sklearn
import sklearn.ensemble
import sklearn.metrics
In [22]:
# fetching data
from sklearn.datasets import fetch_20newsgroups
categories = ['alt.atheism', 'soc.religion.christian']
newsgroups_train = fetch_20newsgroups(subset='train', categories=categories)
newsgroups_test = fetch_20newsgroups(subset='test', categories=categories)
class_names = ['atheism', 'christian']
In [23]:
# vectorizer
vectorizer = sklearn.feature_extraction.text.TfidfVectorizer(lowercase=False)
train_vectors = vectorizer.fit_transform(newsgroups_train.data)
test_vectors = vectorizer.transform(newsgroups_test.data)
In [24]:
# fit
rf = sklearn.ensemble.RandomForestClassifier(n_estimators=500)
rf.fit(train_vectors, newsgroups_train.target)
Out[24]:
RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=None, max_features='auto', max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=500, n_jobs=None,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False)
In [25]:
# predict
pred = rf.predict(test_vectors)
sklearn.metrics.f1_score(newsgroups_test.target, pred, average='binary')
Out[25]:
0.9198606271777003
In [26]:
# lime
from lime import lime_text
from sklearn.pipeline import make_pipeline
c = make_pipeline(vectorizer, rf)
In [27]:
# lime_text
from lime.lime_text import LimeTextExplainer
explainer = LimeTextExplainer(class_names=class_names)
In [28]:
# explain_instance
idx = 83
exp = explainer.explain_instance(newsgroups_test.data[idx], c.predict_proba, num_features=6)
print('Document id: %d' % idx)
print('Probability(christian) =', c.predict_proba([newsgroups_test.data[idx]])[0,1])
print('True class: %s' % class_names[newsgroups_test.target[idx]])

Document id: 83
Probability(christian) = 0.456
True class: atheism
In [29]:
exp.as_list()
Out[29]:
[('Posting', -0.15144490637088556),
 ('Host', -0.14676338634466327),
 ('NNTP', -0.09967779429335712),
 ('edu', -0.01958746993761563),
 ('have', -0.011096783762022664),
 ('From', 0.009432614305286118)]
In [30]:
print('Original prediction:', rf.predict_proba(test_vectors[idx])[0,1])
tmp = test_vectors[idx].copy()
tmp[0,vectorizer.vocabulary_['Posting']] = 0
tmp[0,vectorizer.vocabulary_['Host']] = 0
print('Prediction removing some features:', rf.predict_proba(tmp)[0,1])
print('Difference:', rf.predict_proba(tmp)[0,1] - rf.predict_proba(test_vectors[idx])[0,1])

Original prediction: 0.456
Prediction removing some features: 0.72
Difference: 0.26399999999999996
In [31]:
%matplotlib inline
fig = exp.as_pyplot_figure()
In [32]:
exp.show_in_notebook(text=False)
In [ ]:
# exp.save_to_file('/tmp/oi.html')
In [33]:
exp.show_in_notebook(text=True)
In [34]:
# package
# from __future__ import print_function
import sklearn
import sklearn.datasets
import sklearn.ensemble
import numpy as np
import lime
import lime.lime_tabular
np.random.seed(1)
In [35]:
# iris
iris = sklearn.datasets.load_iris()
In [36]:
# train_test_split
train, test, labels_train, labels_test = sklearn.model_selection.train_test_split(iris.data, iris.target, train_size=0.80)

/mfw_data/algo/rec/anaconda3/lib/python3.7/site-packages/sklearn/model_selection/_split.py:2179: FutureWarning: From version 0.21, test_size will always complement train_size unless both are specified.
  FutureWarning)
In [37]:
# fit
rf = sklearn.ensemble.RandomForestClassifier(n_estimators=500)
rf.fit(train, labels_train)
Out[37]:
RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=None, max_features='auto', max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=500, n_jobs=None,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False)
In [38]:
sklearn.metrics.accuracy_score(labels_test, rf.predict(test))
Out[38]:
0.9666666666666667
In [39]:
# explainer
explainer = lime.lime_tabular.LimeTabularExplainer(train, feature_names=iris.feature_names, class_names=iris.target_names, discretize_continuous=True)
In [40]:
# explain_instance
i = np.random.randint(0, test.shape[0])
exp = explainer.explain_instance(test[i], rf.predict_proba, num_features=2, top_labels=1)
In [41]:
# show
exp.show_in_notebook(show_table=True, show_all=False)
In [42]:
feature_index = lambda x: iris.feature_names.index(x)
In [47]:
print('Increasing petal width')
temp = test[i].copy()
print('P(setosa) before:', rf.predict_proba(temp.reshape(1,-1))[0,0])
temp[feature_index('petal width (cm)')] = 1.5
print('P(setosa) after:', rf.predict_proba(temp.reshape(1,-1))[0,0])
print ()
print('Increasing petal length')
temp = test[i].copy()
print('P(setosa) before:', rf.predict_proba(temp.reshape(1,-1))[0,0])
temp[feature_index('petal length (cm)')] = 3.5
print('P(setosa) after:', rf.predict_proba(temp.reshape(1,-1))[0,0])
print()
print('Increasing both')
temp = test[i].copy()
print('P(setosa) before:', rf.predict_proba(temp.reshape(1,-1))[0,0])
temp[feature_index('petal width (cm)')] = 1.5
temp[feature_index('petal length (cm)')] = 3.5
print('P(setosa) after:', rf.predict_proba(temp.reshape(1,-1))[0,0])

Increasing petal width
P(setosa) before: 0.0
P(setosa) after: 0.0

Increasing petal length
P(setosa) before: 0.0
P(setosa) after: 0.0

Increasing both
P(setosa) before: 0.0
P(setosa) after: 0.0
In [44]:
exp.show_in_notebook(show_table=True, show_all=True)
In [48]:
feature_names = ["Age", "Workclass", "fnlwgt", "Education", "Education-Num", "Marital Status","Occupation", "Relationship", "Race", "Sex", "Capital Gain", "Capital Loss","Hours per week", "Country"]
In [50]:
data = np.genfromtxt('adult.data', delimiter=', ', dtype=str)
In [51]:
labels = data[:,14]
le= sklearn.preprocessing.LabelEncoder()
le.fit(labels)
labels = le.transform(labels)
class_names = le.classes_
data = data[:,:-1]
In [52]:
categorical_features = [1, 3, 5, 6, 7, 8, 9, 13]
In [53]:
categorical_names = {}
for feature in categorical_features:
    le = sklearn.preprocessing.LabelEncoder()
    le.fit(data[:, feature])
    data[:, feature] = le.transform(data[:, feature])
    categorical_names[feature] = le.classes_
In [54]:
data = data.astype(float)
In [55]:
encoder = sklearn.preprocessing.OneHotEncoder(categorical_features=categorical_features)
In [57]:
np.random.seed(1)
train, test, labels_train, labels_test = sklearn.model_selection.train_test_split(data, labels, train_size=0.80)

/mfw_data/algo/rec/anaconda3/lib/python3.7/site-packages/sklearn/model_selection/_split.py:2179: FutureWarning: From version 0.21, test_size will always complement train_size unless both are specified.
  FutureWarning)
In [58]:
encoder.fit(data)
encoded_train = encoder.transform(train)

/mfw_data/algo/rec/anaconda3/lib/python3.7/site-packages/sklearn/preprocessing/_encoders.py:371: FutureWarning: The handling of integer data will change in version 0.22. Currently, the categories are determined based on the range [0, max(values)], while in the future they will be determined based on the unique values.
If you want the future behaviour and silence this warning, you can specify "categories='auto'".
In case you used a LabelEncoder before this OneHotEncoder to convert the categories to integers, then you can now use the OneHotEncoder directly.
  warnings.warn(msg, FutureWarning)
/mfw_data/algo/rec/anaconda3/lib/python3.7/site-packages/sklearn/preprocessing/_encoders.py:392: DeprecationWarning: The 'categorical_features' keyword is deprecated in version 0.20 and will be removed in 0.22. You can use the ColumnTransformer instead.
  "use the ColumnTransformer instead.", DeprecationWarning)
In [59]:
import xgboost
gbtree = xgboost.XGBClassifier(n_estimators=300, max_depth=5)
gbtree.fit(encoded_train, labels_train)
Out[59]:
XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,
       colsample_bytree=1, gamma=0, learning_rate=0.1, max_delta_step=0,
       max_depth=5, min_child_weight=1, missing=None, n_estimators=300,
       n_jobs=1, nthread=None, objective='binary:logistic', random_state=0,
       reg_alpha=0, reg_lambda=1, scale_pos_weight=1, seed=None,
       silent=True, subsample=1)
In [60]:
sklearn.metrics.accuracy_score(labels_test, gbtree.predict(encoder.transform(test)))
Out[60]:
0.8724090280976509
In [61]:
predict_fn = lambda x: gbtree.predict_proba(encoder.transform(x)).astype(float)
In [62]:
explainer = lime.lime_tabular.LimeTabularExplainer(train ,feature_names = feature_names,class_names=class_names,
                                                   categorical_features=categorical_features, 
                                                   categorical_names=categorical_names, kernel_width=3)
In [63]:
np.random.seed(1)
i = 1653
exp = explainer.explain_instance(test[i], predict_fn, num_features=5)
exp.show_in_notebook(show_all=False)
In [64]:
i = 10
exp = explainer.explain_instance(test[i], predict_fn, num_features=5)
exp.show_in_notebook(show_all=False)
In [ ]: